GitHub adds repository-level Copilot metrics for coding agents

GitHub has made repository-level Copilot usage metrics generally available through the Copilot usage metrics REST API. Enterprise and organization reports can now show daily, per-repository pull request activity for Copilot coding agent and Copilot code review.

This is a narrow API change with a broad management consequence. Until now, many teams could see Copilot adoption at the user or organization layer, but they had a weaker view of where agent-assisted pull requests and AI code review activity were actually landing inside the codebase.

Two endpoints expose daily repository reports

GitHub's July 17 changelog adds two one-day report endpoints: one under enterprise reports and one under organization reports. Both use the repos-1-day report path and accept a day=YYYY-MM-DD query parameter.

The reports cover repositories with activity on the selected day. GitHub's documentation says the repository report includes pull request metrics for Copilot Coding Agent and Copilot Code Review, with one entry per repository. The response returns signed download links for report files rather than embedding the full dataset directly in the initial response.

For enterprise reporting pipelines, that shape matters. Teams will need a scheduled job that requests the report, downloads the signed files before they expire, normalizes repository identifiers, and joins the result with internal ownership or service catalog data.

Copilot governance moves from seat counts to codebase impact

Repository-level reporting makes Copilot measurement less abstract. A manager can now ask which repositories receive agent-generated pull requests, where Copilot code review is active, and whether those areas match the rollout plan.

That helps with three practical decisions:

• Target enablement at repositories where agent pull requests are useful but underused.
• Separate high-volume Copilot activity from high-value Copilot activity by joining reports with merge, incident, and review data.
• Spot repositories that should have stricter policy because they handle sensitive systems, regulated code, or deployment-critical paths.

The feature is especially relevant for organizations adopting AI coding agents rather than only inline completion. Agent pull requests create a stronger need for repository ownership, review policy, and audit evidence.

Access depends on Copilot metrics policy and permissions

The new reports are not open to every developer by default. GitHub says enterprise owners, billing managers, organization owners, and custom roles with the relevant "View Copilot Metrics" permission can access them. The Copilot usage metrics policy must also be enabled.

The docs add token-level details. Enterprise endpoints require an appropriate enterprise Copilot metrics permission for fine-grained GitHub App installation tokens, while OAuth or classic personal access token paths need scopes such as manage_billing:copilot or read:enterprise for enterprise reports. Organization endpoints have their own organization-level permission and scope requirements.

Admins should treat the report as governance data. It can reveal where coding agents and review automation are active, so access should match the people responsible for engineering metrics, security review, finance, or platform operations.

What to wire into dashboards now

The immediate implementation step is not a new executive slide. It is a reliable data path. Pull the one-day repository report into an internal warehouse, join it to repository owners, and track changes over time by repository group.

Useful views include daily repositories with Copilot coding agent pull requests, repositories reviewed by Copilot code review, suggestion counts by comment type, and gaps between enabled repositories and actual activity. Pair those metrics with human review outcomes before using them for performance claims.

For teams already measuring AI development workflows, this update gives GitHub Copilot a cleaner evidence layer. It will not prove productivity by itself, but it reduces the guesswork around where agentic coding work is happening.

Sources and methodology

The primary source is GitHub's July 17 changelog entry. GitHub's REST API documentation confirms the enterprise and organization repository report endpoints, permission requirements, report format, and daily report behavior. daily.dev mirrors the changelog item as independent discovery corroboration.